Ground vehicle localization based on inertial measurement signals

Terrain profiles can be a basis of localization even in the absence of Global Positioning System (GPS). Especially, the pitch rate signals of a ground vehicle well represent the fingerprint of the current position. In this paper, we introduce two possible methods for localization. Both methods are based on spectrograms of the signals. The first one is based on hash labels that are widely used for audio searching algorithms. The second is a normalized gray-level correlation method that considers the spectrograms as images and uses an inner product operation to compare the images.